rag 0.1.0

A Rust library and CLI for Retrieval-Augmented Generation
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290

use rag::{
    vector_store::{Document, InMemoryVectorStore, MetadataFilter, MinimalVectorDB, VectorStore},
    DistanceMetric, Index, FlatIndex,
};

// ============================================================
// Tests validating patterns from examples/batch_search.rs
// ============================================================

#[tokio::test]
async fn test_batch_search_ranking_consistency() {
    let store = InMemoryVectorStore::new();

    let docs = vec![
        Document::new("Rust".to_string()).with_embedding(vec![1.0, 0.0, 0.0, 0.0]),
        Document::new("Go".to_string()).with_embedding(vec![0.9, 0.1, 0.0, 0.0]),
        Document::new("Python".to_string()).with_embedding(vec![0.0, 1.0, 0.0, 0.0]),
        Document::new("JavaScript".to_string()).with_embedding(vec![0.0, 0.0, 1.0, 0.0]),
        Document::new("TypeScript".to_string()).with_embedding(vec![0.0, 0.1, 0.9, 0.0]),
    ];
    store.add_batch(docs).await.unwrap();

    let queries = vec![
        vec![1.0, 0.0, 0.0, 0.0],
        vec![0.0, 0.0, 1.0, 0.0],
        vec![0.5, 0.5, 0.0, 0.0],
    ];

    let results = store.search_batch(&queries, 2).await.unwrap();
    assert_eq!(results.len(), 3);

    // Query 1: should find Rust and Go
    assert_eq!(results[0].len(), 2);
    assert_eq!(results[0][0].document.content, "Rust");
    assert_eq!(results[0][1].document.content, "Go");
    assert!(results[0][0].score >= results[0][1].score);

    // Query 2: should find JavaScript and TypeScript
    assert_eq!(results[1].len(), 2);
    assert_eq!(results[1][0].document.content, "JavaScript");
    assert_eq!(results[1][1].document.content, "TypeScript");

    // Query 3: should find Rust and Python (mixed)
    assert_eq!(results[2].len(), 2);
    assert!(results[2][0].score >= results[2][1].score);
}

#[tokio::test]
async fn test_batch_search_euclidean_spatial() {
    let store = InMemoryVectorStore::with_metric(DistanceMetric::Euclidean);

    let docs = vec![
        Document::new("Point A".to_string()).with_embedding(vec![0.0, 0.0, 0.0]),
        Document::new("Point B".to_string()).with_embedding(vec![0.1, 0.1, 0.1]),
        Document::new("Point C".to_string()).with_embedding(vec![10.0, 10.0, 10.0]),
    ];
    store.add_batch(docs).await.unwrap();

    let queries = vec![vec![0.0, 0.0, 0.0], vec![5.0, 5.0, 5.0]];
    let results = store.search_batch(&queries, 2).await.unwrap();

    // Query at origin should find Point A closest, then Point B
    assert_eq!(results[0][0].document.content, "Point A");
    assert_eq!(results[0][1].document.content, "Point B");

    // Query at [5,5,5] should find Point C (at [10,10,10], dist ~8.66) vs Point B (dist ~8.48)
    // Actually Point B is closer: sqrt((5-0.1)^2*3) ~ 8.48, Point C: sqrt(5^2*3) ~ 8.66
    assert_eq!(results[1][0].document.content, "Point B");
}

// ============================================================
// Tests validating patterns from examples/distance_metrics.rs
// ============================================================

#[tokio::test]
async fn test_distance_metrics_produce_different_rankings() {
    let docs = vec![
        Document::new("A".to_string()).with_embedding(vec![1.0, 0.0, 0.0]),
        Document::new("B".to_string()).with_embedding(vec![0.0, 1.0, 0.0]),
        Document::new("C".to_string()).with_embedding(vec![1.0, 1.0, 0.0]),
        Document::new("D".to_string()).with_embedding(vec![0.9, 0.1, 0.0]),
    ];
    let query = vec![1.0, 0.0, 0.0];

    // Cosine should rank D very close to A (similar direction)
    let cosine_store = InMemoryVectorStore::with_metric(DistanceMetric::Cosine);
    cosine_store.add_batch(docs.clone()).await.unwrap();
    let cosine_results = cosine_store.search(&query, 4).await.unwrap();
    assert_eq!(cosine_results[0].document.content, "A");
    // D has similar direction, should be 2nd
    assert_eq!(cosine_results[1].document.content, "D");

    // Dot product: A=[1,0,0] dot [1,0,0]=1, D=[0.9,0.1,0] dot [1,0,0]=0.9, C=1, B=0
    // A=1, C=1 (tie -- depends on DashMap iteration order), D=0.9, B=0
    let dot_store = InMemoryVectorStore::with_metric(DistanceMetric::DotProduct);
    dot_store.add_batch(docs.clone()).await.unwrap();
    let dot_results = dot_store.search(&query, 4).await.unwrap();
    // First result should be either A or C (both score 1.0), we just check it's not B
    assert!(dot_results[0].score > 0.99);
    assert_ne!(dot_results[3].document.content, "A");
    assert_ne!(dot_results[3].document.content, "C");

    // Euclidean: A at distance 0, D at distance 0.1, C at distance 1, B at distance sqrt(2)
    let euclid_store = InMemoryVectorStore::with_metric(DistanceMetric::Euclidean);
    euclid_store.add_batch(docs.clone()).await.unwrap();
    let euclid_results = euclid_store.search(&query, 4).await.unwrap();
    assert_eq!(euclid_results[0].document.content, "A");
    assert_eq!(euclid_results[1].document.content, "D");
    assert_eq!(euclid_results[2].document.content, "C");
}

#[tokio::test]
async fn test_cosine_ignores_magnitude() {
    let store = InMemoryVectorStore::with_metric(DistanceMetric::Cosine);

    let docs = vec![
        Document::new("Small".to_string()).with_embedding(vec![1.0, 0.0]),
        Document::new("Large".to_string()).with_embedding(vec![100.0, 0.0]),
        Document::new("Orthogonal".to_string()).with_embedding(vec![0.0, 1.0]),
    ];
    store.add_batch(docs).await.unwrap();

    let results = store.search(&[1.0, 0.0], 3).await.unwrap();
    // Cosine should see Small and Large as equally similar (same direction)
    assert!(results[0].score > 0.99); // Small
    assert!(results[1].score > 0.99); // Large (same direction)
    assert!(results[2].score < 0.01); // Orthogonal
}

// ============================================================
// Tests validating patterns from examples/pure_memory_rag.rs
// ============================================================

#[tokio::test]
async fn test_pure_memory_rag_document_lifecycle() {
    let store = InMemoryVectorStore::with_metric(DistanceMetric::Cosine);

    let docs = vec![
        Document::new("Rust doc".to_string())
            .with_embedding(vec![0.95, 0.10, 0.05, 0.00])
            .with_metadata("topic".to_string(), "programming".to_string()),
        Document::new("Python doc".to_string())
            .with_embedding(vec![0.10, 0.90, 0.20, 0.05])
            .with_metadata("topic".to_string(), "programming".to_string()),
        Document::new("ML doc".to_string())
            .with_embedding(vec![0.15, 0.20, 0.85, 0.10])
            .with_metadata("topic".to_string(), "ai".to_string()),
        Document::new("Deep learning doc".to_string())
            .with_embedding(vec![0.20, 0.25, 0.80, 0.15])
            .with_metadata("topic".to_string(), "ai".to_string()),
        Document::new("Docker doc".to_string())
            .with_embedding(vec![0.30, 0.40, 0.10, 0.85])
            .with_metadata("topic".to_string(), "devops".to_string()),
    ];

    store.add_batch(docs).await.unwrap();
    assert_eq!(store.count().await.unwrap(), 5);

    // Semantic search for programming
    let rust_results = store.search(&[0.9, 0.1, 0.0, 0.0], 2).await.unwrap();
    assert_eq!(rust_results[0].document.content, "Rust doc");

    // Semantic search for AI
    let ai_results = store.search(&[0.1, 0.1, 0.9, 0.1], 2).await.unwrap();
    assert_eq!(ai_results[0].document.content, "ML doc");

    // Batch query
    let batch_queries = vec![
        vec![0.9, 0.1, 0.0, 0.0],
        vec![0.1, 0.1, 0.9, 0.1],
        vec![0.3, 0.3, 0.1, 0.8],
    ];
    let batch_results = store.search_batch(&batch_queries, 2).await.unwrap();
    assert_eq!(batch_results.len(), 3);
    assert_eq!(batch_results[0][0].document.content, "Rust doc");
    assert_eq!(batch_results[1][0].document.content, "ML doc");
    assert_eq!(batch_results[2][0].document.content, "Docker doc");

    // Document lifecycle
    let all_docs = store.list(10, 0).await.unwrap();
    assert_eq!(all_docs.len(), 5);

    store.delete(&all_docs[0].id).await.unwrap();
    assert_eq!(store.count().await.unwrap(), 4);

    store.clear().await.unwrap();
    assert_eq!(store.count().await.unwrap(), 0);
}

#[tokio::test]
async fn test_metadata_filter_with_embedding_search() {
    let store = InMemoryVectorStore::new();

    let docs = vec![
        Document::new("Rust programming".to_string())
            .with_embedding(vec![1.0, 0.0, 0.0])
            .with_metadata("lang".to_string(), "rust".to_string()),
        Document::new("Rust web framework".to_string())
            .with_embedding(vec![0.9, 0.1, 0.0])
            .with_metadata("lang".to_string(), "rust".to_string()),
        Document::new("Python data science".to_string())
            .with_embedding(vec![0.0, 1.0, 0.0])
            .with_metadata("lang".to_string(), "python".to_string()),
    ];
    store.add_batch(docs).await.unwrap();

    // Search all returns Rust docs first
    let all_results = store.search(&[1.0, 0.0, 0.0], 3).await.unwrap();
    assert_eq!(all_results.len(), 3);

    // Filtered search only returns rust docs
    let filter = MetadataFilter::new().add("lang".to_string(), "rust".to_string());
    let filtered_results = store.search_with_filter(&[1.0, 0.0, 0.0], 3, &filter).await.unwrap();
    assert_eq!(filtered_results.len(), 2);
    assert!(filtered_results.iter().all(|r| r.document.metadata.get("lang") == Some(&"rust".to_string())));
}

// ============================================================
// Tests validating MinimalVectorDB equivalency
// ============================================================

#[tokio::test]
async fn test_minimal_db_equivalent_behavior() {
    let inmem = InMemoryVectorStore::new();
    let minimal = MinimalVectorDB::new();

    let docs = vec![
        Document::new("Doc 1".to_string()).with_embedding(vec![1.0, 0.0]),
        Document::new("Doc 2".to_string()).with_embedding(vec![0.0, 1.0]),
    ];

    inmem.add_batch(docs.clone()).await.unwrap();
    minimal.add_batch(docs.clone()).await.unwrap();

    let inmem_results = inmem.search(&[1.0, 0.0], 2).await.unwrap();
    let minimal_results = minimal.search(&[1.0, 0.0], 2).await.unwrap();

    assert_eq!(inmem_results.len(), minimal_results.len());
    assert_eq!(inmem_results[0].document.content, minimal_results[0].document.content);
    assert_eq!(inmem_results[1].document.content, minimal_results[1].document.content);

    assert_eq!(inmem.count().await.unwrap(), minimal.count().await.unwrap());

    inmem.clear().await.unwrap();
    minimal.clear().await.unwrap();
    assert_eq!(inmem.count().await.unwrap(), 0);
    assert_eq!(minimal.count().await.unwrap(), 0);
}

// ============================================================
// Tests validating FlatIndex directly (from Index trait)
// ============================================================

#[test]
fn test_flat_index_metric_accessor() {
    let index_cosine = FlatIndex::new();
    assert_eq!(index_cosine.metric(), DistanceMetric::Cosine);

    let index_euclid = FlatIndex::with_metric(DistanceMetric::Euclidean);
    assert_eq!(index_euclid.metric(), DistanceMetric::Euclidean);
}

#[test]
fn test_flat_index_batch_search_parallel() {
    let index = FlatIndex::new();
    index.add(Document::new("A".to_string()).with_embedding(vec![1.0, 0.0]));
    index.add(Document::new("B".to_string()).with_embedding(vec![0.0, 1.0]));
    index.add(Document::new("C".to_string()).with_embedding(vec![0.5, 0.5]));

    let queries = vec![
        vec![1.0, 0.0],
        vec![0.0, 1.0],
        vec![0.5, 0.5],
    ];

    let results = index.search_batch(&queries, 2);
    assert_eq!(results.len(), 3);

    // Each query should return 2 results
    for r in &results {
        assert_eq!(r.len(), 2);
    }

    // Query 1 should find A first
    assert_eq!(results[0][0].document.content, "A");
    // Query 2 should find B first
    assert_eq!(results[1][0].document.content, "B");
    // Query 3 should find C first (perfect match)
    assert_eq!(results[2][0].document.content, "C");
}